The Trump administration has asked the Center for AI Standards and Innovation to stop issuing public reports on AI models, people familiar with the matter said, a move that tightens government control over model evaluation as national security concerns escalate. CAISI, housed within the Commerce Department, is the primary government body that tests advanced AI models before their public release and publishes findings about their capabilities and relative performance. Officials including National Cyber Director Sean Cairncross told CAISI to halt publication of its assessments while an executive order President Trump signed last week is implemented, the people said.
The order was a victory for Cairncross and Treasury Secretary Scott Bessent, who have argued that security agencies should play a larger role in model evaluation, according to people familiar with the matter. The directive put CAISI’s future in question. The unit is still working internally to evaluate models and coordinate with government agencies, but has stopped public-facing work, the people said.
Some administration officials have expressed frustration with Cairncross for exerting more influence over the process, the people said. Those officials believed the executive order directed a new group to do work CAISI was already performing.
White House spokeswoman Liz Huston said in a statement that “the implementation of President Trump’s AI agenda is a whole-of-government effort, with numerous agencies contributing to its success.” She said Cairncross “is doing excellent work to protect the American people while also promoting innovation.”
On the other side of the debate, companies including OpenAI have held discussions with administration officials about the importance of CAISI and preserving its power, the people said. White House AI adviser and venture capitalist David Sacks has warned that an overzealous model-testing process — regardless of which agency oversees it — could slow deployment and hinder innovation.
CAISI was originally established as the AI Safety Institute under President Biden. Commerce Secretary Howard Lutnick renamed and restructured the unit at the start of Trump’s second term. In April, a former Anthropic researcher who had been leading CAISI for only days was asked to resign by the White House, a move some officials believed stemmed from the administration’s feud with the AI developer.
Top model developers including OpenAI and Anthropic have maintained relationships with CAISI since the Biden administration. The process outlined in the executive order could complicate how companies plan to release models if they fear national-security officials might deem them risky and slow deployment, AI analysts said.
Last week, OpenAI called for CAISI to be strengthened with more resources and responsibilities in a policy blueprint. Chris Lehane, the company’s chief global affairs officer, told reporters that the unit was capable of sophisticated testing and had established relationships with leading companies.
AI experts said CAISI is underfinanced compared to similar government bodies in other countries and is needed to maintain U.S. leadership in the field. The move to restrict its activity is the latest example of how powerful models such as Anthropic’s Mythos — capable of carrying out cyberattacks — have forced the administration to reconsider its hands-off AI approach, analysts said.